dyyyyyyyy/SCAN-Base
Viewer
• Updated
• 101k • 11
SCAN
We propose Self-Denoising Monte Carlo Annotation (SCAN), an efficient Process Reward Model (PRM) data synthesis and noise-tolerant learning framework.
We propose Self-Denoising Monte Carlo Annotation (SCAN), an efficient Process Reward Model (PRM) data synthesis and noise-tolerant learning framework.